Predicting Health Care Cost Transitions Using a Multidimensional Adaptive Prediction Process.

Area of Science:

• Health Services Research
• Health Informatics
• Predictive Analytics

Background:

• Population health management requires balancing individual care needs with efficiency and quality.
• Predictive models can assess prospective risk for targeted health management.
• Accurate cost prediction is crucial for resource allocation in healthcare.

Purpose of the Study:

• To develop and test the Multidimensional Adaptive Prediction Process (MAPP) for predicting future healthcare costs.
• To compare MAPP's predictive accuracy against a status quo reference model.

Main Methods:

• MAPP divides populations into cost cohorts.
• It uses multiple models and covariates for optimized cost prediction within each cohort.
• Tested on 3 years of administrative claims data for members with coronary heart disease.

Main Results:

• MAPP identified members contributing $7.9M and $9.7M more in 2011 costs for increasing/high-cost cohorts, respectively.
• MAPP achieved an $1882 per member annual improvement, a 21% increase in accurate cost capture.
• Demonstrated improved prospective cost prediction compared to the reference model.

Conclusions:

• A novel adaptive multiple-model approach enhances future cost prediction accuracy.
• MAPP enables efficient resource allocation by identifying individuals with increasing costs.
• Improved prospective identification supports better care quality for emergent needs.